93 results about "Partial least squares regression" patented technology

Disclosed is a method and device for predicting soil organic carbon based on spectrum-guided ensemble learning. The method includes: obtaining a soil sample and a real organic carbon content and an original soil spectrum thereof, pre-processing the original soil spectrum to obtain a soil spectrum sample, constructing, based on the soil spectrum sample, a sample set, grouping the sample set into a first training set and a validation set, and using the real organic carbon content as a label; training, based on the first training set and the corresponding labels, a partial least squares regression model, a Cubist model and a random forest model to obtain carbon content predicted value sets of the three models; constructing, based on the carbon content predicted value sets of the three models and soil spectrum principal component data, a second training set, and training a second random forest model with the second training set and corresponding labels to obtain a spectrum-guided ensemble model. The method combines the advantages of different predictive models and can accurately predict the soil carbon content.

The invention discloses a distillate oil property prediction method based on deep learning feature extraction and partial least squares regression. The method comprises the following steps: firstly, carrying out classification training on a near infrared spectrum through a convolution-attention double-branch fusion network, and extracting high-dimensional spectral features with local and global information; then, historical samples are retrieved from a database based on prediction categories, a plurality of most similar samples are selected by adopting cosine similarity measurement to construct a correction set, and the spectral features and property labels are subjected to standardization processing; and finally, carrying out partial least squares regression modeling on the correction set, extracting latent variables to maximize covariance between spectral features and physicochemical properties, and inputting feature vectors of an oil sample to be detected into the trained PLS model to obtain a corresponding property prediction result. According to the method, the modeling requirement and the category specificity characteristics under the small sample condition are considered while the prediction precision is guaranteed, and the method is suitable for rapid property detection and intelligent analysis in the refining process.

The invention relates to the technical field of laser cladding repair, and provides a positioning and clamping system and method for repairing the surface of a large working roll of a steel mill based on laser cladding, a three-field coupling prediction model of a temperature field, a stress field and a displacement field is established, and a partial least squares regression algorithm is combined with a Kalman filter to realize multi-step prospective prediction. According to the predicted thermal deformation, self-adaptive compensation control is implemented through a three-layer cascade compensation mechanism composed of a hydraulic drive, a piezoelectric ceramic driver and a laser head adjusting mechanism. The system adopts a time domain, frequency domain and time-frequency domain joint analysis method to extract vibration characteristics, vibration types are classified and identified through a support vector machine, and an active-passive hybrid suppression strategy is adopted for different vibration sources. High-precision positioning and stable clamping of the large working roller in the laser cladding repairing process are achieved, and the thickness uniformity and the surface quality of a cladding layer are remarkably improved.

The invention discloses a method for detecting the nitrogen content of organ tissues of different varieties of oilseed rapes based on visible near-infrared spectroscopy, which comprises the following steps: firstly, collecting oilseed rape leaf, shell, stalk and root system samples under the conditions of multiple varieties and multiple nitrogen fertilizer levels, and acquiring spectral data within the range of 430-2500nm by using a visible near-infrared spectroscopy; and a Kjeldahl method is synchronously adopted to measure the real nitrogen content as a reference value. Preprocessing the spectral data, including de-noising, standard normal variable transformation, multivariate scatter correction and derivative processing, so as to weaken the influence of scattering and baseline drift; characteristic wavelengths related to the nitrogen content are screened through stepwise regression, variable projection importance, competitive self-adaptive reweighted sampling, a continuous projection algorithm and other methods, and a random forest model, a support vector machine model, a partial least squares discriminant analysis model, a partial least squares regression model, a support vector regression model, an XGBoost model and other models are combined. And respectively constructing a classification identification model and a regression prediction model.

The invention discloses a wheat yield remote sensing prediction method combining phenological parameters, which comprises the following steps: S1, performing field observation in a key growth period of winter wheat, synchronously collecting canopy hyperspectral reflectivity data, leaf area index (LAI) and SPAD value of each observation sample point, and recording wheat grain yield of the corresponding sample point; s2, performing noise reduction preprocessing on the acquired canopy hyperspectral data, and extracting sensitive spectral parameters by combining principal component analysis (PCA) and correlation analysis methods; s3, taking the sensitive spectrum parameters, LAI and SPAD values as independent variables, taking the wheat grain yield as a dependent variable, and adopting partial least squares regression (PLSR) to construct a multivariable yield prediction model; and S4, performing wheat yield prediction on an independent test sample or regional scale remote sensing data by using the trained model. The method overcomes the defect that only yield sensitive spectrum parameters are used for predicting the effect, and accurate estimation of the model is achieved.

The invention relates to the technical field of agricultural remote sensing and precise irrigation, in particular to a method and equipment for inverting leaf water potential of grapes before dawn based on multispectral remote sensing, and the method comprises the steps: measuring leaf water potential of representative grape plants in each planting test plot before dawn in different growth periods, and collecting canopy multispectral images in the noon period; preprocessing the remote sensing original image data; calculating a vegetation index; on the basis of the model fitting data set and the training feature set, constructing basic prediction models in different growth periods by adopting a unary linear regression algorithm, a multiple linear regression algorithm and a partial least square regression algorithm respectively; determining a PLSR inversion model according to the decision coefficient and the root-mean-square error; and carrying out pre-dawn leaf water potential prediction through the PLSR inversion model. Through the method, high-precision inversion of the water potential of the leaves before dawn based on the noon canopy multispectral data can be realized, and reliable technical support is provided for accurate management of vineyard moisture and intelligent irrigation.

The invention relates to a method for detecting the INDF content in TMR of lactating cattle based on a near infrared spectrum. The method comprises the following steps: constructing a near infrared spectrum model capable of being used for detecting the INDF content; wherein a sample set for constructing the near infrared spectrum model comprises near infrared spectrum data corresponding to a lactating cattle TMR sample and an INDF actual value; the method comprises the following steps: extracting a wavelength point with a non-zero coefficient in near infrared spectrum data, and taking the wavelength point as a characteristic variable; constructing a near infrared spectrum model based on the characteristic variables and the corresponding INDF actual values by adopting a partial least squares regression algorithm; model parameters are optimized through a cross validation method, an optimal model parameter combination is determined by combining a cross validation root mean square error RMSECV and an external validation root mean square error RMSEP, and a near infrared spectrum model capable of being used for INDF content detection is obtained; and inputting near infrared spectrum data of a to-be-detected lactating cattle TMR sample into the near infrared spectrum model to obtain an INDF content detection result.

The invention relates to the technical field of leather wastewater treatment, and provides an ultraviolet-visible spectrum-based leather wastewater ozone dosage and COD (Chemical Oxygen Demand) prediction method. The method comprises the following steps: firstly, determining the types of organic matters in a tannery wastewater sample, and determining the distribution characteristics of the organic matters in different fluorescent regions through three-dimensional fluorescence spectrum analysis, so as to discriminate the types of organic pollutants; secondly, screening key wavelengths, and determining characteristic absorption wavelengths representing different organic matters through an ultraviolet-visible light spectrum; then carrying out an ozone oxidation experiment, and synchronously measuring a COD value and an ultraviolet absorption spectrum under different ozone addition amounts; based on experimental data, a multiple linear regression and exponential regression prediction model of ozone dosage and a COD prediction model based on partial least squares regression are established. According to the invention, accurate control of ozone dosage and real-time monitoring of COD are realized, the treatment efficiency is improved, the operation cost is reduced, the adaptability to water quality fluctuation is enhanced, and the method has a wide engineering application prospect.

The present application provides a kind of biogas slurry chemical oxygen demand online monitoring method and system, the method comprises: obtaining the current characteristic data of biogas slurry, the characteristic data includes multiple in dissolved oxygen, turbidity, oxidation-reduction potential and conductivity;The current characteristic data of the biogas slurry is input into least square regression model, and the current chemical oxygen demand predicted value of the biogas slurry output by the least square regression model is obtained;The least square regression model is obtained by training by taking the characteristic data sample of the biogas slurry as characteristic variable, and taking the measured value of the chemical oxygen demand corresponding to the biogas slurry as target variable.The present application realizes the rapid, low-cost detection of biogas slurry chemical oxygen demand by real-time monitoring of multiple characteristic data, constructing multi-parameter fusion partial least square regression prediction model.

The invention relates to a structural network-genetic map biological network model for predicting ischemic stroke and a construction method thereof, and the method comprises the steps: extracting and calculating seven multi-scale morphological features and pairwise Pearson correlation coefficients among the features from T1 weighted imaging data and diffusion tensor imaging data; constructing a 308 * 308 morphological similarity network matrix and a brain network module for identifying ischemic stroke neural dysfunction; 1782 sampling points are extracted from the Airy human brain map, and each sampling point comprises expression data of 10185 genes; the method comprises the following steps: mapping space coordinates of AHBA sampling points to a cortex package of a Desikan-Killiany map, carrying out normalization processing to output 308 * 10185 brain region gene-by-gene expression matrixes, and constructing a structural network-gene map biological network model for predicting ischemic stroke by adopting a partial least square regression method and a bootstrap method. Compared with the prior art, the model determines the specific molecular mechanism related to the phenotypic structure change of ischemic stroke injury, and the stroke occurrence probability is predicted according to the specific molecular mechanism.

The invention belongs to the technical field of insulator detection, and discloses an on-line non-contact detection method for the density of sand attached to the surface of an insulator, and the method comprises the following steps: S1, obtaining a hyperspectral line of an insulating sheet with known sand attached density; s2, establishing a parameter detection model of the sediment density, and solving by adopting a partial least square regression method; and S3, carrying out online non-contact detection on the sand-attached density of the insulator. According to the online non-contact detection method for the density of the sand attached to the surface of the insulator, the problems that an existing detection method needs power-off operation, the detection process is complex, deviation is prone to being generated, and the density of the local sand attached cannot be accurately reflected are solved; on-line non-contact detection is carried out on the surface sand density of the insulator in the power transmission line under the condition that power is not cut off, operation is easy and convenient, the influence of the sand attaching state on the insulation performance of the insulator can be accurately evaluated, and a reliable test basis is provided for insulation design and maintenance of the insulator.

The application provides a night image quality evaluation method. First, different scales of night images are obtained through downsampling, and then the original image and the image sequence after downsampling are simultaneously sent into a feature extraction unit; in the feature extraction stage, two types of quality-related features, namely, bottom visual features and high-level semantic features, are extracted. The bottom visual features include texture features and contrast features of the night image; for the high-level semantic features, first, a deep convolution network is used to extract primary high-level semantic features of the night image, and then mean and variance two feature functions are used to aggregate the primary high-level semantic features to obtain secondary high-level semantic features as the final high-level semantic features of the night image. Then, the extracted bottom features and high-level features are aggregated. Finally, a partial least squares regression method is used to regress the aggregated features, so as to obtain an objective score of the night image quality.

The application discloses an evaporation dish evaporation amount homogenization sequence reconstruction method based on multi-source data fusion, specifically, daily evaporation dish evaporation amount data of each station in a target area, meteorological data and station metadata are screened, automatic secondary quality control is performed on the screened data, and a high-quality basic data set is obtained; a random forest regression and recursive feature elimination cross-validation method is used for selecting features, and finally a partial least squares regression model is used for splicing and interpolation on the daily evaporation dish evaporation amount data, so that the evaporation dish evaporation amount reconstruction sequence is generated; the evaporation dish evaporation amount reconstruction sequence is subjected to homogenization test, breakpoints are identified and corrected, and the evaporation dish evaporation amount homogenization reconstruction sequence is obtained, so that the problem of discontinuous evaporation amount data sequence caused by non-climatic factors such as mixed use of large and small evaporation dishes is solved, and the limitations of traditional conversion coefficient method and multivariate regression equation, such as large difference between stations, insufficient consideration of meteorological conditions and the like, are overcome.

The invention discloses a method and system for constructing a drinking water taste quantitative evaluation model, and belongs to the technical field of drinking water quality evaluation, and the method comprises the steps: S1, collecting a plurality of drinking water samples; s2, performing standardized sensory evaluation to obtain a comprehensive taste score of each drinking water sample; s3, performing water quality analysis to obtain water quality detection data; s4, based on the taste comprehensive score and the water quality detection data, screening out a preset number of key water quality indexes from multiple water quality indexes by adopting a variable importance projection method; and S5, taking the key water quality index as an independent variable, taking the taste comprehensive score as a dependent variable, and adopting a partial least square regression method to construct a drinking water taste quantitative evaluation model. The method solves the problem that the existing drinking water taste evaluation can only depend on subjective evaluation of a large number of testers and is poor in uncertainty and timeliness. The method is objective in quantification, accurate in evaluation, high in guidance and convenient to apply, and can realize high-frequency daily taste evaluation.

This invention relates to the field of food component detection technology, specifically a method for detecting the content of components in sesame paste based on spectral analysis. The method includes: collecting near-infrared transmission spectral data of a target sesame paste sample at different temperature gradients to construct a temperature-compensated spectral feature matrix; inputting this matrix into a multi-task partial least squares regression model to simultaneously predict the content of protein, fat, and carbohydrates; comparing the predicted values ​​with a spectral library of sesame paste standard substances, and fine-tuning the model's latent variable space to converge the fitting error; recalculating the corrected predicted values ​​using the fine-tuned model, and fusing them with the theoretical values ​​of the formula to generate a comprehensive component content estimation vector; finally, generating a detection result record containing the specific values ​​of each target component. This method can improve the accuracy and efficiency of sesame paste component detection and is suitable for sesame paste food quality testing.

The application belongs to the technical field of food safety early warning, and particularly relates to a method for predicting heavy metal content in table grapes, comprising the following steps: using a partial least squares regression method to analyze the correlation between soil physicochemical indexes and grape heavy metal content, and retaining indexes with a VIP value greater than 1 as key variables; taking the key variables as input variables and grape detection results as output to construct a random forest regression model; training the random forest regression model through repeated execution of a sampling and training process to obtain a prediction model of heavy metals in grapes; and predicting the heavy metal content in table grapes according to the generated prediction model of heavy metals in grapes.

The invention belongs to the technical field of bletilla striata polysaccharide preparation, and provides a high-purity bletilla striata polysaccharide preparation quality control method, which comprises: constructing a raw material spectrum-polysaccharide content prediction model by using Fourier transform near infrared spectroscopy and partial least squares regression PLSR, and intelligently determining an initial material-liquid ratio based on a predicted value; in the extraction stage, the concentration is monitored in real time through an online refractometer, the ultrasonic power, the extraction time or the volume of a supplemented solvent is dynamically adjusted, and it is ensured that the concentration and the viscosity of an extracting solution are in the optimal interval; in the purification stage, the dosage of a purification reagent and the heating rate are dynamically adjusted according to the real-time concentration; in the concentration stage, according to the real-time concentration and viscosity, the vacuum degree is adjusted in a segmented mode, heating is controlled, and end point coking is prevented; according to the invention, on-line monitoring and closed-loop feedback control of key parameters in the whole process are realized, and the purity, uniformity and production stability of the product are remarkably improved.

This invention relates to the field of traditional Chinese medicine component research technology, specifically to a method for constructing fingerprint spectra based on the antioxidant active components of Ganoderma lucidum and its application. First, a method for ultrasonic-assisted acid hydrolysis of crude Ganoderma lucidum polysaccharides guided by free radical scavenging activity was constructed. This method is efficient, easy to operate, and can effectively reflect the free radical scavenging activity of the depolymerized Ganoderma lucidum polysaccharides. Hydrophilic chromatography-electrospray ionization detector-electrospray mass spectrometry was used to acid hydrolyze and analyze 52 batches of Ganoderma lucidum polysaccharides from different sources. Further, combined with grey relational analysis and partial least squares regression analysis, 12 common activity peaks were screened, establishing fingerprint spectra based on the antioxidant active components of Ganoderma lucidum. Finally, principal component analysis and partial least squares discriminant analysis were used to analyze the fingerprint spectra based on the antioxidant active components of Ganoderma lucidum for standard Ganoderma lucidum, Ganoderma sinense, Ganoderma applanatum, Ganoderma lucidum var. truncatum, and sample Ganoderma lucidum, respectively, achieving the identification of Ganoderma lucidum varieties.

The present disclosure relates to a method for predicting the enthalpy of a phase change material, a near-infrared spectrum of a phase change material sample is subjected to a partial least squares regression analysis with a standard enthalpy to establish a first calibration model; then a spectral fitting residual matrix is selected according to the construction process of the first calibration model, the spectral fitting residual matrix and the enthalpy matrix are combined, and are used as input signals and teacher signals of an extreme learning machine respectively to obtain a second calibration model; through the combined application of the first calibration model and the second calibration model, the linear relationship and the nonlinear relationship between the near-infrared spectrum of the phase change material sample and the enthalpy can be comprehensively considered, the enthalpy prediction capability in a wide range is achieved, and the model prediction accuracy is improved.

The application discloses a pesticide detection method for rice based on QuEChERS and ultraviolet-visible spectrophotometry, and takes three pesticides, i.e., bensulfuron-methyl, propanil and cypermethrin, as examples to carry out research on a rapid screening method for multiple pesticide residues in rice. The improved QuEChERS method is used to process rice to obtain a matrix liquid, matrix standard solution and solvent standard solution are prepared, the quantitative linear range of each pesticide and the matrix effect are obtained. In the quantitative concentration range, mixed solutions of the three pesticides in four combinations are prepared, the types of the pesticides in the rice are determined through support vector machine qualitative analysis, and the concentrations of the pesticides are predicted through partial least squares regression analysis, so that the pesticide residues in the rice can be effectively and rapidly detected.

The application discloses an electrical equipment sound recognition method based on feature frequency band selection. The method comprises the following steps: sound time domain data acquisition and preprocessing, conversion to a frequency domain by using Fourier transform; a Metropolis-Hastings sampling method is used to randomly sample a set number of samples to establish a partial least squares regression model; a logarithmic attenuation function is used to filter the frequency band with the minimum absolute value weight of the regression coefficient; a set retention ratio of the frequency band is selected by using weighted sampling, and the corresponding frequency band with a smaller root mean square error of cross-validation is selected as a feature frequency band; a neural network is established to recognize the sound state. The application reduces the spatial dimension of the training sample, avoids data redundancy, and improves the training speed. The application effectively recognizes three states of normal working, normal air leakage and abnormal air leakage, and improves the recognition accuracy.

The present application provides a kind of total nitrogen concentration on-line monitoring method and system in biogas slurry, the method comprises: obtaining the current characteristic data of biogas slurry, the characteristic data includes multiple in dissolved oxygen, turbidity, oxidation-reduction potential and conductivity;The current characteristic data of the biogas slurry is input into least square regression model, and the current total nitrogen concentration prediction value of the biogas slurry output by the least square regression model is obtained;The least square regression model is obtained by training by taking the characteristic data sample of the biogas slurry as characteristic variable, and taking the total nitrogen concentration measured value corresponding to the biogas slurry as target variable.The present application realizes the rapid, low-cost detection of total nitrogen concentration of biogas slurry by real-time monitoring of multiple characteristic data, constructing multi-parameter fusion partial least square regression prediction model.

The application discloses a soil salinization discrimination method based on combination optimization of geochemical parameters, and belongs to the field of soil salinization monitoring. The method comprises the following steps: collecting soil samples in a research area and measuring the content of multi-dimensional geochemical elements; removing outliers and performing standardization processing on the geochemical data; dividing the samples by using a clustering algorithm; screening characteristic elements which are significantly related to salinization and have low collinearity in each division to form an optimal parameter combination; establishing a partial least squares regression model based on the combination, constructing a salinization response function and calculating a discrimination index value; training and optimizing the index by using a machine learning model to obtain a discrimination model; and finally, realizing rapid discrimination of the degree of soil salinization by using the model. The application considers regional spatial heterogeneity and multi-element coupling characteristics, has the advantages of high discrimination accuracy and good stability, and is suitable for regional scale salinization investigation, grading evaluation and early warning.

The application discloses a multi-particle size combination optimized optical fingerprint silt source tracing method, which comprises the following steps: dividing silt source types into S type and G type, collecting corresponding type samples, wet screening the samples into three particle size ranges respectively, and measuring the spectra; performing PCA-LDA discriminant analysis on the samples obtained after wet screening based on the spectra; preparing a series of mixed samples by different combination modes and different gradient proportions according to the samples through discriminant analysis, and measuring the spectral data for modeling; performing modeling by using a partial least squares regression method, fitting the original spectral data and the spectral data after spectral pretreatment of the modeling set samples with the source sample proportion values of the corresponding samples respectively, establishing a spectral tracing model, and screening the best tracing model from the spectral tracing model; and predicting and evaluating silt sources by using the best tracing model. The application can improve the accuracy of optical fingerprint tracing results.

The invention discloses a polyester fiber analysis method and system based on infrared spectroscopy, and relates to the technical field, and the method comprises the following steps: obtaining a PET sample; carrying out sample analysis on the PET sample, and collecting infrared spectrum data; extracting peak intensity or area characteristic data of a plurality of characteristic peaks in the corresponding infrared spectrum data; and constructing a PET regeneration analysis model, and inputting the feature data into the PET regeneration analysis model to obtain qualitative and quantitative analysis results. Four key characteristic peaks highly related to the regeneration process in a PET spectrum are screened as a minimum characteristic set, and partial least squares discriminant analysis and partial least squares regression are combined, so that rapid identification and content quantification of regenerated PET are realized. The method does not need an expensive chemical analysis platform and complex sample pretreatment, is high in detection speed, is wide in applicable sample form, remarkably reduces the detection cost, improves the detection flux, and meets the requirements for regenerated PET proportion verification, tracing and quality control in spinning, packaging and supervision scenes.

The present application relates to the technical field of food nutrition component detection and information automatic processing, and particularly discloses a kind of food nutrition component rapid metering detection and label generation system.The present application obtains original spectrum data by adopting preset near-infrared wave band detection light beam and is preprocessed after multivariate scattering correction and first derivative transformation, inputs the calibration model established based on partial least square regression algorithm to output the metering value of multiple nutrition component contents, then completes numerical rounding according to preset rounding rule and generates nutrition label page data by matching and filling with label template mark field, solves the technical problems that traditional laboratory chemical analysis method detection cycle is long, pre-treatment is complicated, equipment dependency is strong and detection data and label generation link lack of automatic connection, leading to manual input error-prone, realizes the synchronous improvement of food nutrition component detection speed, label generation efficiency and data accuracy.

The application provides a kind of insulator contamination hyperspectral feature extraction detection method and system, belong to power equipment detection technical field, including: to insulator sample is scanned to multiple wave bands, obtain the original hyperspectral image of preset wave band, and original hyperspectral image is corrected using black and white correction algorithm, based on the spectral response characteristics of insulator sample contamination component, according to the signal-to-noise ratio and spectral difference degree of each wave band of corrected hyperspectral image, filter out characteristic wave band data;Using first derivative transformation to pre-process characteristic wave band data;From the pre-processed characteristic wave band data, extract multi-dimensional feature parameters;Construct the fusion prediction model based on support vector machine and partial least squares regression, input multi-dimensional feature parameters into fusion prediction model, output the contamination grade and equivalent salt density value of insulator sample.The application improves the accuracy of insulator sample contamination hyperspectral detection through data quality optimization, efficient feature screening, integrated model construction and other designs.

The invention belongs to the technical field of data processing, and particularly relates to a peptide milk powder activity quality detection method based on wet process production, which comprises the following steps: S1, acquiring fluid time sequence data of a production node and an original reflection spectrum sequence of a spray drying node, performing smooth filtering operation on the original reflection spectrum sequence to obtain a standardized smooth spectrum; s2, importing the smooth spectrum into a pre-configured partial least square regression algorithm model, and calculating and outputting the apparent chemical concentration of the peptide component in the milk powder of the current batch; and S3, based on background features of the smooth spectrum and fluid mechanics features of the fluid time sequence data, constructing an operation index with causal logic to complete depth correction operation on the apparent chemical concentration, and outputting real activity retention concentration with both concentration attribute and activity attribute. According to the invention, the optical measurement error is effectively eliminated, and the real biological value is accurately evaluated.

The present application relates to a kind of methods for screening anti-inflammatory, antioxidant active compounds from natural Chinese herbal medicine wormwood based on spectrum-effect relationship, the method establishes the method for the common characteristic peak of the fingerprint chromatogram of wormwood extract ultra-high performance liquid chromatography.By grey correlation analysis, Pearson bivariate correlation analysis method and partial least square regression analysis etc. chemometrics method establishes the spectrum-effect relationship of wormwood anti-inflammatory, antioxidant activity, to screen the anti-inflammatory, antioxidant active compounds in wormwood.The analysis result shows that compound chlorogenic acid, isochlorogenic acid A, isochlorogenic acid C is the main anti-inflammatory, antioxidant active compound in wormwood extract.The present application provides scientific and effective method for the research of the medicinal material basis of wormwood and quality control, and provides reference for the development and utilization of wormwood.

Systems and methods are disclosed for performing fault detection and prediction for power electronics and switching devices for power electronics, such as power inverters. Systems and methods disclosed herein can include determining, by a partial least squares model that evaluates values for one or more switching parameters for a switching device, the one or more switching parameters selected from a first set of switching parameters, a predicted value for the on-state current Ids of the switching device. The predicted value for the on-state current Ids can be based on the values of the one or more switching parameters for the switching device. Systems and methods disclosed herein can determine a residual comprising the difference between the predicted value for the another switching parameter of the switching device and an actual value of the predicted value for the another switching parameter, and generate a test statistic based on the residual.