Chromatogram Peak Separation Using a Learned Discriminator
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional peak detection methods in chromatography require operator intervention for setting detection parameters, leading to inaccuracies and time-consuming trial-and-error processes, especially when dealing with unseparated peaks.
Innovation Solution
A method involving the generation of a discriminator through machine learning using unseparated waveform data, where peaks of different compositions are superimposed to generate training data, allowing for accurate peak detection and separation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional peak detection methods are used, then operator skills and manual parameter setting are required, but detection accuracy decreases and time consumption increases
Solution Approach 1:
The system performs self-learning by automatically generating training data from measured chromatogram data through superimposition processing. The discriminator learns peak detection parameters autonomously without operator intervention, enabling the system to serve itself in optimizing detection accuracy while eliminating time-consuming manual parameter setting
Solution Approach 2:
The discriminator is trained in advance using learning data generated from superimposed chromatogram peaks. This preliminary learning phase enables the system to automatically detect peaks with high accuracy during actual measurement without requiring real-time operator intervention or trial-and-error parameter adjustment
2Measurement precision
If conventional peak detection algorithms are used, then simple parameter setting is possible, but accurate detection in unseparated peaks cannot be achieved
Solution Approach 1:
A discriminator acting as an intermediary component is introduced between the chromatogram data and peak detection output. This discriminator, trained on superimposed peak data, mediates the detection process by automatically resolving unseparated peaks, thereby achieving high detection accuracy without increasing operational complexity
Solution Approach 2:
The manual mechanical process of operator-based parameter setting and trial-and-error detection is replaced with an automated machine learning system. The discriminator automatically processes chromatogram data and detects peaks, substituting human expertise with an autonomous algorithmic system that handles complex unseparated peak scenarios
Data Source
AI summary
An object is to accurately detect peaks of various compositions, even in a case of unseparated peaks in which peaks of a plurality of compositions are superimposed. A computer acquires waveform data D1 having a peak P1 in a composition A measured by a data analysis device (S10). Next, the computer acquires waveform data D2 having a peak P2 in a composition B measured by the data analysis device (S20). Next, waveform data D12 including unseparated peaks by superimposing the waveform data D1 including the acquired peak P1 and the waveform data D2 including the acquired peak P2 (S30) is generated. Next, the generated waveform data D12 of the unseparated peaks is input as learning data, and the waveform data D1 and D2 corresponding to the waveform data D12 are input as training data in Step S40. Next, machine learning is performed using the waveform data D12, D1, and D2, and a learned model for estimating an accurate separation method of unseparated peaks is constructed based on the trained result (S50).