Synthetic Multicomponent Samples for Calibration Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for developing multivariate calibrations and supervision systems for industrial processes face challenges due to complex multicomponent matrices, low acquisition frequency, and high resource consumption, particularly in processes like fermentations, where dynamic changes and high correlation coefficients between parameters complicate the development of accurate and robust calibrations.
Innovation Solution
A method utilizing Multivariate Curve Resolution - Alternating Least Squares (MCR-ALS) to create synthetic multicomponent samples that mimic dynamic processes, allowing for the generation of historical process data, determination of necessary main solutions, and creation of augmented time-dependent matrices to enhance calibration accuracy and robustness, reducing the need for extensive offline analytical measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If process samples are used for developing multivariate calibrations, then the calibration reflects real process conditions, but the method requires high resource consumption and time for extensive offline analytical measurements
Solution Approach 1:
The patent creates synthetic process samples that replicate the spectral characteristics and chemical composition variations of real process samples without requiring actual process runs. This copying approach maintains calibration reliability while eliminating the time-consuming nature of collecting and analyzing real process samples through offline methods
Solution Approach 2:
The patent performs preliminary analysis of historical process data to identify key spectral features and composition patterns before creating synthetic samples. This preliminary action enables the synthesis process to focus on reproducing only the critical variations needed for accurate calibration, reducing overall development time
2Measurement precision
If orthogonal designs of experiments are used to produce non-process derived samples, then the calibration meets spectral selectivity and variability requisites, but the calibration results are comparably poor for spectroscopy sensors
Solution Approach 1:
The patent transforms the approach by changing from orthogonal experimental designs to synthetic sample generation based on historical process data. This parameter change allows the synthetic samples to reflect actual process spectral characteristics and composition relationships, improving calibration accuracy while maintaining spectral selectivity through controlled variation of key parameters
Solution Approach 2:
The patent introduces synthetic samples as an intermediary between orthogonal experimental designs and real process samples. These synthetic samples serve as a mediator that combines the controlled variability of experimental designs with the realistic spectral characteristics of actual process data, achieving both selectivity and accuracy
3Adaptability or versatility
If real process samples are collected for calibration development, then unknown and non-measurable variations are present, but processes can take several days (for instance fermentations) and require high number of samples
Solution Approach 1:
The patent creates synthetic copies of process samples that capture the full range of process variations including unknown and non-measurable components. By copying the spectral and compositional characteristics from historical data, the system achieves comprehensive process coverage without requiring actual multi-day process runs
Solution Approach 2:
The patent performs preliminary extraction and analysis of process variations from historical data before synthetic sample generation. This preliminary action identifies and preserves all critical variations including unknown components, enabling comprehensive coverage to be achieved computationally rather than through extended physical processes
Data Source
Figure 1
Figure 1
Figure 1
AI summary
The present invention describes a method that is comprised of creating a set of synthetic samples that mimic a dynamic process or a specific process step or a variation thereof, to be used to develop multivariate monitoring calibrations and process supervisory control systems. In order to enhance calibration models accuracy and robustness the introduction of randomly spiked, programed spiked, random mixing and orthogonal experimental designed samples is also described. The new method works by using prior knowledge of dynamic processes expressed by their available data and quality control data and using such data to build mixtures profiles that can be used to build synthetic samples that mimic the process dynamics. A novel methodology to overcome calibration developments drawbacks of the prior art can be provided comprising designed synthetic multicomponent samples that mimic dynamic processes and using orthogonal design of experiments or random spikes or programed spikes or random mixing to enhance calibration models performance. With such multicomponent synthetic-samples which approximate accurately the matrix main components composition of real-samples we have demonstrated a novel use of PAT monitoring tools to estimate the dynamic process state (process supervision). The capability to derive multivariate process trajectories with multiparametric analysis of synthetic-samples prepared to span the multivariate design-space of a process allows for process supervisory control strategies to be established during process development and optimization.