Fig. 1. Schematic of sensing acquisition system based on thulium-doped fiber ring cavity

Fig. 2. Physical map of sensing acquisition system based on thulium-doped fiber ring cavity

Fig. 3. Output spectra of thulium-doped fiber ring cavity system

Fig. 4. Sample diagrams of absorption spectrum data. (a)-(c) Absorption spectrum data after pretreatment of pure NH₃ gas with volume fraction of 2%, absorption spectrum data after Lorentz fitting, and absorption spectrum data obtained by simulation according to HITRAN database; (d)-(f) absorption spectrum data after pretreatment of pure CO₂ gas with volume fraction of 5%, absorption spectrum data after Lorentz fitting, and absorption spectrum data obtained by simulation according to HITRAN database; (g)-(i) pre-processed absorption spectrum data of mixed gas of NH₃ with volume fraction of 2% and CO₂ with volume fraction of 2%, absorption spectrum data after Lorentz fitting, and absorption spectrum data obtained by HITRAN database simulation

Fig. 5. Partial principal component distributions after PCA dimensionality reduction on training set. (a) NH₃-SVM model; (b) CO₂-SVM model

Fig. 6. Flowchart of adaptive mutation particle swarm optimization algorithm

Fig. 7. Error curves of adaptive particle swarm optimization for model train set. (a) NH₃-SVM model; (b) CO₂-SVM model

Fig. 8. Comparison of results of train set and test set. (a) Comparison of set volume fraction and predicted volume fraction for NH₃-SVM model train set; (b) comparison of set volume fraction and predicted volume fraction for CO₂-SVM model train set; (c) comparison of set volume fraction and predicted volume fraction for NH₃-SVM model test set; (d) comparison of set volume fraction and predicted volume fraction for CO₂-SVM model test set