1 Introduksi
1.1 Latar Belakang
Electronic nose (e-nose) dan electronic tongue (e-tongue) adalah sistem sensor array yang meniru kemampuan penciuman dan pengecapan manusia melalui sensor silang-reaktif multi-sensor dan algoritme pengenalan pola (Persaud & Dodd, 1982; Toko, 2000; Loutfi et al., 2015). E-nose umumnya memakai sensor metal oxide semiconductor (MOS), polimer konduktif, atau quartz crystal microbalance (QCM) untuk senyawa aroma volatil. E-tongue memakai sensor potentiometric, voltammetric, atau ion-selective untuk senyawa terlarut. Jembatan ke ilmu sensori perlu dibuat karena sinyal e-nose/e-tongue bukan intensitas atribut secara langsung, melainkan sidik-jari pola yang dihubungkan dengan profil deskriptif melalui PLSR atau PCA bridge (Ross, 2009; Vlasov et al., 2005). E-nose / E-tongue Bridge menyusun matriks sensor × sampel, mereduksi dimensi melalui PCA, dan memprediksi deskriptor sensori dari pola sensor.
1.2 Tujuan Modul
- Menerima data sensor array dalam bentuk matriks sampel × sensor.
- Menjalankan PCA fingerprinting untuk identifikasi clustering sampel, autentikasi, dan konsistensi batch.
- Mendukung model prediktif PLSR: pola sensor → target descriptor.
- Mengukur akurasi klasifikasi untuk tugas identifikasi berbasis LDA.
- Audiens: peneliti dengan instrumen e-nose/e-tongue dan tim R&D yang membutuhkan QC sensori cepat.
1.3 Posisi di Antara Alternatif
Pilih E-nose / E-tongue Bridge untuk analisis multivariat sensor array. Untuk prapemrosesan NIR, gunakan NIR Preprocessing Explorer. Untuk PLSR umum, gunakan PLSR Studio. Untuk eksplorasi PCA, gunakan PCA Explorer (segera hadir). Untuk klasifikasi terawasi, gunakan Class Modelling.
2 Metode
2.1 Dasar Teoretis
Sensor array data structure:
$$X_{n \times p} = \begin{pmatrix} x_{11} & \cdots & x_{1p} \\ \vdots & \ddots & \vdots \\ x_{n1} & \cdots & x_{np} \end{pmatrix}$$dengan $n$ samples dan $p$ sensors (mis. 8 MOS sensor untuk e-nose, 7 ISFET sensor untuk e-tongue).
Pre-processing standard (Yan et al., 2015):
- Baseline correction — subtract pre-exposure baseline per sensor.
- Feature extraction — use $R_{\max}$, area under response curve (AUC), or initial slope.
- Autoscaling — mean-center + unit variance per sensor.
PCA:
$$X = T P^T + E$$dengan $T$ scores (sample positions) dan $P$ loadings (sensor contributions). Sample clustering di PC space identifies batch consistency atau outlier detection.
PLSR sensor → descriptor:
$$y_{\text{descriptor}} = X B_{\text{PLS}} + \varepsilon$$dengan $B_{\text{PLS}}$ regression coefficient dari NIPALS. Output: prediction descriptor target from sensor pattern + VIP score for sensor importance.
LDA for classification (Fisher, 1936):
$$\text{Maximize} \quad \frac{|S_B|}{|S_W|}$$dengan $S_B$ between-class scatter dan $S_W$ within-class scatter. Output: linear combinations of sensors that are maximally discriminative for class labels.
Cross-validation accuracy:
$$\text{Accuracy} = \frac{\text{Correct predictions}}{\text{Total predictions}}$$Target for e-nose authentication: > 90%.
2.2 Persamaan Inti
- PCA decomposition: $X = T P^T + E$
- PLSR prediction: $\hat{y} = X B_{\text{PLS}}$
- LDA optimization: $\max |S_B|/|S_W|$
- Cross-validation accuracy: correct/total
2.3 Asumsi & Batas Validitas
| Asumsi | Konsekuensi jika dilanggar | Cara cek di SQalytics |
|---|---|---|
| Sensor pre-conditioning konsisten | Drift hari-ke-hari | Calibration check per sesi |
| Replikasi $\geq 3$ per sampel | Variance bias | Modul flag |
| Sample preparation SOP | Volatile loss inconsistent | Catat protokol |
| Sensor not poisoned (degraded) | Sensitivity drop | Reference standar mingguan |
| PCA / PLSR variance captured ≥ 70% | Underfit | Cek explained variance |
| Class balanced untuk LDA | Bias predominant class | Modul flag imbalance |
3 Cara Kerja
3.1 Step-by-Step di SQalytics
- Buka
E-nose / E-tongue Bridgedari domain Sensori dan Riset Konsumen. - Muat sensor matrix: kolom
Sample+ multi-sensor columns (mis. S1, S2, ..., S8). - (Opsional) Muat descriptor target untuk PLSR (matrix Y).
- Pre-process: autoscaling, baseline correction, feature extraction.
- Pilih analysis mode:
PCA Fingerprinting,PLSR Sensor → Descriptor,LDA Classification. - Klik Run Analysis.
- Tinjau: Tab
PCA Biplot, TabPLSR Prediction, TabLDA Classification.
3.2 Template Tabel Input + Contoh Data Sintetis
Input template: Sample (category ✓), Class (category — opsional untuk LDA), S1–Sn (numeric ✓).
| Sample | Class | S1 | S2 | S3 | S4 | S5 | S6 | S7 | S8 |
|---|---|---|---|---|---|---|---|---|---|
| K001 | Aceh | 250 | 380 | 145 | 520 | 290 | 410 | 180 | 320 |
| K002 | Aceh | 245 | 385 | 142 | 525 | 285 | 415 | 178 | 318 |
| K003 | Toraja | 290 | 320 | 180 | 480 | 340 | 380 | 210 | 290 |
| K004 | Toraja | 295 | 318 | 178 | 478 | 342 | 385 | 215 | 285 |
| K005 | Gayo | 235 | 410 | 130 | 555 | 270 | 440 | 165 | 350 |
| K006 | Gayo | 232 | 415 | 128 | 558 | 268 | 442 | 168 | 348 |
| … (30 samples total: 10 per class) | |||||||||
docs/assets/example-data/id/sensory/template_sensory_e_nose_bridge.csv.
3.3 Contoh Luaran
PCA result — class centroids di PC space:
| Class | PC1 mean | PC2 mean |
|---|---|---|
| Aceh | +0.85 | −0.40 |
| Toraja | −1.20 | +0.60 |
| Gayo | +1.50 | −0.15 |
LDA confusion matrix (leave-one-out CV):
| Predicted Aceh | Predicted Toraja | Predicted Gayo | |
|---|---|---|---|
| Actual Aceh (n=10) | 10 | 0 | 0 |
| Actual Toraja (n=10) | 0 | 10 | 0 |
| Actual Gayo (n=10) | 1 | 0 | 9 |
Accuracy = 29/30 = 96.7%
PLSR Studio bila ingin model sensor → descriptor (mis. predict cupping score), atau ke Class Modelling (SIMCA) untuk one-class authentication yang lebih strict.4 Kesimpulan
4.1 Relevansi Real-World
- Coffee origin authentication — discriminasi asal geografis dari profil aroma volatil.
- Honey adulteration detection — deteksi adulterant via e-tongue sugar profile.
- Wine quality grading — rapid classification tier dari sensor array.
- Tea grading — discriminasi kualitas (TGFOP vs BOP) tanpa panel manusia.
- Spoilage detection — meat/dairy freshness monitoring via volatil pattern.
- Process monitoring fermentation — real-time QC profil aroma selama fermentasi.
4.2 Where to Go from Here
⚙ Troubleshooting Cepat
i Riwayat Revisi
| Tanggal | Revisi | Penulis |
|---|---|---|
| 2026-05-12 | Draft v2 publikasi (KaTeX PCA/PLSR/LDA sensor array + APA Persaud/Toko/Vlasov/Loutfi) | Claude |
4 Referensi
- Persaud, K., & Dodd, G. (1982). Analysis of discrimination mechanisms in the mammalian olfactory system. Nature, 299(5881), 352–355. https://doi.org/10.1038/299352a0
- Toko, K. (2000). Biomimetic sensor technology. Cambridge University Press.
- Loutfi, A., Coradeschi, S., Mani, G. K., Shankar, P., & Rayappan, J. B. B. (2015). Electronic noses for food quality: A review. Journal of Food Engineering, 144, 103–111. https://doi.org/10.1016/j.jfoodeng.2014.07.019
- Vlasov, Y., Legin, A., Rudnitskaya, A., Di Natale, C., & D'Amico, A. (2005). Nonspecific sensor arrays ("electronic tongue"). Pure and Applied Chemistry, 77(11), 1965–1983. https://doi.org/10.1351/pac200577111965
- Ross, C. F. (2009). Sensory science at the human-machine interface. Trends in Food Science & Technology, 20(2), 63–72. https://doi.org/10.1016/j.tifs.2008.11.004
- Yan, J., Guo, X., Duan, S., Jia, P., Wang, L., Peng, C., & Zhang, S. (2015). Electronic nose feature extraction methods: A review. Sensors, 15(11), 27804–27831. https://doi.org/10.3390/s151127804
- Fisher, R. A. (1936). The use of multiple measurements in taxonomic problems. Annals of Eugenics, 7(2), 179–188. https://doi.org/10.1111/j.1469-1809.1936.tb02137.x