Electronic Nose Sensing in Agriculture A Feasibility Study dr. ir. Aleksandar Andreski Dept. of NanoPhysics Interfaces Saxion University of Applied Sciences
Contents Application: detecting rot in potato storage Piezo-microcantilevers as electronic nose Preliminary Results and Future Work Project Participants Partners
Potato Storage Matching supply & demand requires storage on site Storage time ~1 year New harvest is still wet and vulnerable to rot Rot spreads very quickly First few weeks are critical è Monitoring is needed
Potato Storage Systems Warehouses are ventilated with active cooling Control of Temperature (and partially Humidity) C0 2 is also monitored 30% losses without warehousing system Big problem in e.g. developing world
Potato Storage Systems Multiple T and RH Sensors Automatic Recipes Energy Minimization Can Track Batch from Field to Customer
Potato Storage Systems Add one more sensor E-Nose Monitors gas composition in real time Early warning for disease Challenges (specs): Ppm detection limit T and RH compensated Position, sampling?
E-Nose Working Prnciple Human olfactory system e-nose: array of non-specific, cross-reactive sensors combined with an information processing system Many Sensors of Some Specificity Literature examples Complex Analyte Statistical Analysis PC2: 24.5% Principal component analysis 4 Q4 Propanol 2 Cyclohexanon Q6 0 Q3 Q2 Anisole Q7 Q8 Q5-2 Toluene Q1-4 -4-2 0 2 4 PC1: 64.6% 2
Olfactory Detection of Rot Detection of Potato Storage Disease via Gas Analysis: A Pilot Study Using Field Asymmetric Ion Mobility Spectrometry Massimo Rutolo 1, *, James A. Covington 1, John Clarkson 2 and Daciana Iliescu 1 Sensors 2014, 14, 15939-15952; doi:10.3390/s140915939 Many Literature Studies dating back to 1990s Different Types of Rot Different Biomarkers VOC Fingerprinting This line is selected for highest observability. Gas ions mobility spectra Statistical analysis (PCA) and Clustering
Olfactory Detection of Rot Detection of Potato Storage Disease via Gas Analysis: A Pilot Study Using Field Asymmetric Ion Mobility Spectrometry Massimo Rutolo 1, *, James A. Covington 1, John Clarkson 2 and Daciana Iliescu 1 Sensors 2014, 14, 15939-15952; doi:10.3390/s140915939 Many Literature Studies dating back to 1990s Different Types of Rot Different Biomarkers VOC Fingerprinting
Olfactory Detection of Rot The development of a sensor system for the early detection of soft rot in stored potato tubers To cite this article: B P J de Lacy Costello et al 2000 Meas. Sci. Technol. 11 1685 Many Literature Studies dating back to 1990s Different Types of Rot Different Biomarkers VOC Fingerprinting VOC Biomarkers Different positions in a crate
MicroCantilever Principle Sorption changes Mass & Spring constant è Resonant f shift
Δf f = 1 2 Δk k Δm m WATER ACETONE ETHANOL Saturation (%) Saturation (%)
Piezoelectric MEMS cantilever with PZT Layer Silicon SiO 2 Pt Photoresist PZT Insulation Gold Pyralin 200um
SURFACE MASS LIMIT OF DETECTION SMLOD * Thermomechanical Noise Limit * In practice, limit is due to electronic noise A=Displacement amplitude Device Resonance Quality Surface Minimum SMLOD DMMP DMMP area Frequency factor sensitivity Relative Concentration Concentration Device & Ref Frequency resolution resolution deviation (measured) (estimated) (μμ C ) (MHz) (cμ C. l 1 ) (Hl. μμ C ) (ppb) (ppb) FBAR [122] 5 10 4 (b) 1100 210 726 3.6 10 7 (b) 10000 (b) (c) 60 (e) SAW [34] 5 10 7 (b) 158 (c) 100 7 10 8 7000 (b) 87 (b) 42 CMR [123] 6 10 3 (b) 180 5 10 4 2.3 10 3 (b) 1.3 10 8 60 700 (d) 0.35 (e) CMUT [124] 1 10 6 (b) 47.7 140 4.1 10 3 (b) 1.15 10 8 80.5 15 (d) 3 (e) Nanocantilevers [50] μ-cantilevers [125][126] Table 1-1 1.5 3.2 10 3 (b) 10 0.1 200 80 3.75 10 3 (b) 1.5 10 7 2.8 10 3 Paul Ivaldi. Modeling, fabrication and characterization of resonant piezoelectric nano mechanical systems for high resolution chemical sensors. Micro and nanotechnologies/microelectronics. Université de Grenoble, 2014. English. <NNT : 2014GRENT109>. <tel-01192918> (b) 10 8 400 (b) (c) 80 53 25 PZT offers high transduction factors Higher electronic amplitudes Less electronic noise in practice (d) 2 (e) (e)
Readout and Measurements Mechanical resonance Transduced by PZT parasitic Phase-Lock-Loop Resonator Interface + Tracks resonant f automatically + mhz noise with these sensors - Complex to implement in product - Suffers from C P variation Z m Low Piezo-transduction Z m >> Z p weak resonance High Piezo-transduction stonger resonance peaks
Readout and Measurements Mechanical resonance Transduced by PZT parasitic Phase-Lock-Loop Resonator Interface + Tracks resonant f automatically + mhz noise with these sensors - Complex to implement in product - Suffers from C P variation
Readout and Measurements Frequency noise [mhz rms] 100 10 1 RMS readout noise as a function of readout speed Δf f 5 10 8 (50 ppb) 0.1 10 100 1 k PLL Bandwidth [Hz] Δm m eff = 2 Δf f Δm S min 1 ag µm 2 Practical SMLOD (can be lower if we measure slower)
Readout and Measurements: ACETONE 157590 Cantilever C10 (PEI), Flow of 0.1L/min, PLL Target bandwidth = 60Hz 157850 Two Cantilevers Simulatneously, Flow of 50mL/min PLL Target bandwidth 10Hz 160752 157588 0 ppm 0 ppm 0 ppm 157849 160751 157586 250ppm Resonance freq. [Hz] 157584 157582 157580 157578 Shift [mhz] 500 ppm 15 10 5 0-5 -10-15 -20-520 -510-500 -490 time [s] 1000 ppm 11Hz (11mHz/ppm) Resonant frequency [Hz] 157848 157847 157846 157845 ~100ppm ON/OFF Sensor with PEI layer Sensor with PAA layer Noise x10 (unexplained) 160750 160749 160748 160747 1000 ppm 157576-800 -600-400 -200 0 Time [s] 157844 160746-5000 -4000-3000 -2000-1000 0 Time [s]
Conclusions and Future Work Detection of potato rot via E-nose is present in literature and prior work. Piezo-cantilevers with PZT have: a low theoretical SMLOD good practical properties (high signals) Measurements so far are not conclusive Drift is very high and unpredictable Noise increases with more cantilevers active Sensitivity to parasitics with current readout scheme (PLL) Future Work: Sensitive Layers Layer Deposition Readout Scheme
Many Thanks to: Ruud Steenwelle Albert van Hoorn and all students that contributed greatly: Sameh, Bas, Steven, Jorrik, Casper, Ferry. Questions?