Quantum cascade laser (QCL), as an ideal coherent source in mid-infrared wavelength range, has drawn increasing attention in the gas sensing area. Herein, we integrate a tunable QCL source and a hollow core fiber as the gas cell in a compact sensing system, which lead to limit of detection (LOD) of 1.5 ppm for methane. Artificial intelligence (AI), using advanced extreme learning machine (ELM), is then introduced to realize quantitative spectra analysis. This method is superior to using normal linear algorithm model.
TECHNOLOGY FEATURES & SPECIFICATIONS
We have realized system integration of a portable gas monitor, using a hollow core fiber gas cell, which can increase the optical length to 5 meters while maintaining minimized gas volume (
We have developed an innovative ELM-based regression model for spectra analysis of gas sensing. Compared with conventional methods, ELM auto encoder (ELM-AE) can simultaneously achieve dimension reduction and feature extraction, which provide an improved performance compared with conventional algorithms.
We have proposed a tunable QCL array, as the broad tuning range of the laser is beneficial for multi-gas sensing. The high resolution of QCL spectra also helps to distinguish gas mixtures with strong interference.
The broadband tuning range of the QCL slot array and the advanced ELM algorithm help to detect gas mixtures with complex components, for example, food safety analysis, which usually tests various components.
Potential application can be chemical analysis, environmental gas monitoring, food safety and quality check, in-line process monitoring, etc.
The reduced size and weight of our system means that the integration of gas sensors in mobile carriers is more favourable, ensuring long-term monitoring. Our ELM architecture targets to train the output weights to obtain enhanced calculation speeds.