Search

Incorporating monitoring and forecasting technologies has the potential to enhance the different aspects of winemaking. By monitoring certain chemical parameters, it is possible to evaluate the progress of the fermentation and phenolic extraction, while also detecting deviations from expected behaviour. This study aims to investigate the development of a rapid forecasting method and different strategies to improve the accuracy of process monitoring. The first part involves the introduction of partial least squares (PLS) calibration models able to forecast the trajectories of phenolic extraction during fermentation using grape extract spectra. Additionally, novel methods for reoptimizing calibration models are proposed, with the intention of increasing the prediction accuracy. Whilst this method is transferable to PLS models, principal component regression (PCR) models were used instead as a more simplistic way to test the new methods. First, dynamic PCR linear regression coefficients were adapted over time to capture the compound evolution in the wine matrix. Secondly, PCR calibration models were reoptimized using interpolated spectral and reference data. The forecasting PLS calibration models have shown promising results for total anthocyanin content (mg/L), colour density, polymeric pigments (mg/L), MCP tannins (mg/L), and total phenolic index with overall R2 values on the validation set above 0.6, and low RSMEV values. Using interpolated data to re-optimise models was found to be more effective than global models or those using the calculated regression coefficients methods for predicting the concentration of a particular compound at a certain time point. While R2 values are suboptimal for all considered PCR models, in 72% of the cases analysed, the interpolation method for re-optimisation showed RMSEV values between 6% and 83% lower than those reported for a more traditional PCR model and other re-optimisation strategies. Interpolated models have shown consistency in predicting values, along with good performance metrics, indicating that they could be applied to samples for which calibration models do not exist. • This study uses grape extracts spectral data to predict wine phenolic content. • Time-specific calibration models were optimised to increase prediction accuracy. • Acceptable R2 and RMSEP for phenolic prediction in fermentation were obtained. • Using interpolated data to re-optimise models was found to be effective. [ABSTRACT FROM AUTHOR]

Multivariate calibration models are one of the most predominant tools in analytical chemistry for the non-destructive quantification of constituents. While pipelines, methods, and software are widely available for training calibration models, the monitoring strategies to track the prediction performance in time are still limited and are one of the emerging research topics in chemometrics and machine learning. Moreover, for the sake of efficiency, the development of modern monitoring strategies requires tools that do not require new reference values at each time point. The current study addresses the problem of monitoring calibration models for regression. We present a monitoring paradigm and propose two methods to monitor the performance of the model. The two methods focus on bias drift and spectral variation drift. We present the derivation of the methods and their application to two real case studies from the agrofood industry involving milk composition and apple harvesting. Conclusions and recommendations are presented for practitioners, as well as views for future work in this domain of research. • Views and concepts about multivariate calibration monitoring are presented. • Two unsupervised metrics to monitor calibration models in regression were derived. • Focus on monitoring models for potential bias and spectral variation drift. • Two real cases from the agrofood industry were monitored. [ABSTRACT FROM AUTHOR]

Multivariate spectral signals are highly correlated. Often, variable selection techniques are deployed, aiming at model optimization, identification of key variables to explore the underlying physicochemical system or development of a cheap multi-spectral system based on key variables. However, many times the selected variables do not supply a good estimate of properties when tested on a new setting such as new measurements performed on a different spectrometer, different physical or chemical state of the samples and difference in the environmental factors around the experiment. Often the model based on variables selected in the first domain (specific conditions/instrument) does not generalize on the new domain (specific conditions/instrument). To deal with it, in the present work a new method to variable selection called domain invariant covariate selection (di-CovSel) is proposed. The method selects the most informative variables which are invariant to the differences in the instruments, physical or chemical state of the samples and the differences in the environmental factors around the experiment. The method is inspired by domain invariant partial least-square (di-PLS) and the covariate selection (CovSel). The potential of the method is demonstrated on four real cases related to the calibration of near-infrared (NIR) spectroscopy on agri-food materials. The results show that in all the cases, the domain invariant features selected by the di-CovSel have low prediction error compared to the standard variable selection with the CovSel approach when the models are tested on a new data domain. In summary, domain invariant features selected across domains support the development of calibration models with good generalization and supply a better understanding of the system by bypassing the external factors originating from differences in the instruments, physical or chemical states of the samples and the differences in the environmental factors around the experiment. Note that one key feature of the proposed method is that the most important variables which generalize well across domains can be identified without requiring reference measurements in the target domain. • A domain invariant variable selection technique is presented. • Method selects variables to achieve generalized models. • Several application cases for domain invariant variable selection are presented. [ABSTRACT FROM AUTHOR]

Multivariate spectral signals are highly correlated. Often, variable selection techniques are deployed, aiming at model optimization, identification of key variables to explore the underlying physicochemical system or development of a cheap multi-spectral system based on key variables. However, many times the selected variables do not supply a good estimate of properties when tested on a new setting such as new measurements performed on a different spectrometer, different physical or chemical state of the samples and difference in the environmental factors around the experiment. Often the model based on variables selected in the first domain (specific conditions/instrument) does not generalize on the new domain (specific conditions/instrument). To deal with it, in the present work a new method to variable selection called domain invariant covariate selection (di-CovSel) is proposed. The method selects the most informative variables which are invariant to the differences in the instruments, physical or chemical state of the samples and the differences in the environmental factors around the experiment. The method is inspired by domain invariant partial least-square (di-PLS) and the covariate selection (CovSel). The potential of the method is demonstrated on four real cases related to the calibration of near-infrared (NIR) spectroscopy on agri-food materials. The results show that in all the cases, the domain invariant features selected by the di-CovSel have low prediction error compared to the standard variable selection with the CovSel approach when the models are tested on a new data domain. In summary, domain invariant features selected across domains support the development of calibration models with good generalization and supply a better understanding of the system by bypassing the external factors originating from differences in the instruments, physical or chemical states of the samples and the differences in the environmental factors around the experiment. Note that one key feature of the proposed method is that the most important variables which generalize well across domains can be identified without requiring reference measurements in the target domain. • A domain invariant variable selection technique is presented. • Method selects variables to achieve generalized models. • Several application cases for domain invariant variable selection are presented. [ABSTRACT FROM AUTHOR]

Indirect quantification of chemical composition through spectral measurements requires the establishment of multivariate calibration models. The reference analyses on the calibration samples typically form a major cost factor in the establishment of these multivariate models. Therefore, the aim of this study was to select the most informative calibration samples in an unsupervised way based on the spectral measurements. To this end, guidelines to address this challenge in PLSR model building have been developed. The recommendations include calculating a sample size that surpasses the model complexity by a factor of 12, performing the selection in the PCA score space spanned by a sufficiently large number of principal components and using methods such as Kennard-Stone, Puchwein, Clustering or D-optimal designs. We provide the data and methodology used in the present study for future use. • We aim to find a subset that would render a model of comparable performance to the model rendered by the full set of samples. • The optimal sample size for a multivariate calibration model depends on the complexity of such a model. • It is possible to evaluate the quality of the selected samples by comparing covariance matrices for PLSR calibration models. • Control over the sample size and the input dimensionality lowers the impact of the selection method. [ABSTRACT FROM AUTHOR]