Spectrometer Calibration Using Multi-Deflection ML Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional spectrometry methods require expert users to select a single deflection amount for determining analyte concentrations, which limits accuracy and requires discarding valuable information from multiple peaks in the output intensity signal.
Innovation Solution
A spectrometer support system that generates arrays of spectrometer output intensities for calibration samples at different known concentrations, trains a machine-learning computational model to predict analyte concentrations based on these arrays, and uses the model for subsequent spectrometer operations, thereby utilizing more of the output intensity signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single deflection amount is selected for determining analyte concentration, then the operation is simplified, but the measurement precision deteriorates due to discarding valuable information from multiple peaks
Solution Approach 1:
The patent combines multiple deflection amounts (multiple peaks) into a single analytical framework by training a machine learning model to process the entire array of spectrometer output intensities across all deflection amounts simultaneously, merging the information from all peaks to determine analyte concentration
Solution Approach 2:
The patent introduces a machine learning computational model as an intermediary between the raw spectrometer data and the analyte concentration determination, enabling the system to process multiple deflection amounts automatically without requiring expert user intervention to select individual peaks
2Measurement precision
If expert users manually select deflection amounts, then measurement precision may be maintained, but the ease of operation deteriorates and requires specialized knowledge
Solution Approach 1:
The patent enables the spectrometer system to perform calibration and analysis automatically through the machine learning model, which self-adjusts to optimize the use of multiple deflection amounts without requiring expert user intervention or specialized knowledge for peak selection
Solution Approach 2:
The patent transforms the calibration approach by changing from manual parameter selection (expert users choosing specific deflection amounts) to automated parameter optimization (machine learning model processing the entire intensity array across all deflection amounts)
3Device complexity
If only a single peak is used for analysis, then the device complexity is reduced, but the loss of information from multiple peaks decreases productivity and accuracy
Solution Approach 1:
The patent transitions from one-dimensional analysis (single peak at one deflection amount) to multi-dimensional analysis by processing the entire array of spectrometer output intensities across multiple deflection amounts, utilizing the additional dimensional information to improve accuracy without significantly increasing operational complexity
Data Source
AI summary
Disclosed herein are spectrometer support systems, as well as related methods, computing devices, and computer-readable media. For example, in some embodiments, a spectrometer support apparatus may: receive, for each of a plurality of calibration samples of an analyte at different known concentrations, an array of spectrometer output intensities of the calibration sample, wherein different ones of the spectrometer output intensities in the array are associated with data representative of different deflection amounts; train a machine-learning computational model, using the plurality of known concentrations of the analyte in the calibration samples and the associated plurality of arrays of spectrometer output intensities, to output a concentration of the analyte in a target sample based on an input array of spectrometer output intensities of the target sample; and use the trained machine-learning computational model as a calibration model for the analyte for subsequent spectrometer operation.