January 13, 2022 at 12:00 pm | Updated July 14, 2022 at 4:14 pm | 4 min read
The use of precision tools, such as spectrophotometers, to measure fruit quality in farms is rising with the advent of affordable technology. Like any other tool, one could expect that age and improper maintenance might take their toll, affecting the accuracy of quality estimations. The study explored in this article was designed specifically to test this hypothesis. Let’s dive into what they discovered.
What Can Go Wrong in NIR Spectrophotometers?
Researchers use visible and shortwave near-infrared (SWNIR) spectrophotometers to assess quality parameters like total soluble sugars (TSS), dry matter, titrable acidity (TA), and the color of fruits. These parameters are then used as harvest indices and in produce sorting on the farm. Though users might do everything right throughout the testing process, maintenance is not always consistent.
Without proper maintenance, the performance of SWNIR spectrophotometers can decrease due to the following issues:
Subscribe to the Felix instruments Weekly article series.
By submitting this form, you are consenting to receive marketing emails from: . You can revoke your consent to receive emails at any time by using the SafeUnsubscribe® link, found at the bottom of every email. Emails are serviced by Constant Contact
- Light source or halogen bulb aging can lead to lamp failure.
- If the lamp causes an increase in temperature, the device will become more sensitive to longer wavelengths due to silicon photodiode photo-response. Temperature increase can also lead to increased thermal noise.
- Changes can occur in the detector system, such as wavelength drift, spectral sensitivity, and alteration of signal to noise ratio.
Changes in the lamp and detector can impact TSS predictions at shorter wavelengths. Other minute changes can also affect wavelength and probe alignment, altering calibration model performance.
Hence, a group of food scientists, Acharya, Hayes, Subedi, & Walsh, tested how the accuracy of some commonly used commercial tools changed over time. They evaluated changes in spectral quality and their impact on predictive model performance in a long-term experiment over three years.
**NOTE: If you have questions about the maintenance of your Felix Instruments device not found in your manual, or need help addressing a problem you are experiencing, please contact Felix Instruments Support. Our knowledgeable, friendly support staff are always ready to help you get the most out of your device.**
Portable Spectrophotometers Tested
The scientists tested three models of NIRvana, direct predecessors to the Portable NIR Quality Analyzers currently available from Felix Instruments – Applied Food Science.
Data collection: Each of the three tools used a Zeiss MMS1 spectrometer and a halogen lamp as the light source. The scientists tested the devices after periods of intensive field use. Once each year, 24 ripe fruits were tested. They were scanned by the SWNIR spectrophotometers twice on two sides of the fruits. The juice was extracted from a tissue core at the same location the spectra were collected—the TSS measured by a refractometer. In addition, the scientists collected 20 spectra from PTFE white tile as reference.
Chemometrics: Absorbance spectra were collected from the reference tiles and ripe apples at 3nm intervals between 732-936 nm. They were then preprocessed to give the second derivative Savitsky-Golay absorbance spectra used in partial least square regression (PLSR) models to predict TSS. The researchers used Matlab 2014a and PLS toolbox 7.5.1 for the software analysis and employed a full cross-validation model.
Results obtained over three years were analyzed by a Principle component analysis (PCA).
Tools and Prediction Models are Robust
The reference PTFE tiles helped show a difference in spectra measurements by the three instruments in the 400-500 nm. Each device used an internal gold-plated shutter, as gold is a good reflector of infrared light, but absorbs visible light. A variation in these gold reference plates was the reason for the difference in the three instruments’ absorbance spectra of the same white tile.
The twenty repeat measurements of the white tiles allowed the scientists to compare the repeatability of each portable spectrophotometer. They used the standard deviation of the 20 spectral measurements as a measure of the instrument repeatability. Even though the instrument repeatability was relatively stable for all three tools over the course of the experiment, there was one exception. Unit 05 showed poor repeatability in one year.
The difference between tools and years in measurements or lower repeatability is due to differences in the gold coating of the shutters, which was reduced but not entirely removed by considering the second derivative of absorbance spectra (D2A), see Figure 1(a and b).
Apple TSS Predictions are Accurate
The PCA of apple TSS measurements didn’t show much difference over the three years for any of the spectrophotometers used, Figure 1(c). There was no difference between years, and it was also not correlated to the performance repeatability of each device. In each case, the repeatability values were up to 3.5 mAbs and were satisfactory for TSS prediction.
Moreover, the prediction accuracy of the units tested was comparable to results reported in other studies for apple.
PLSR models are robust enough to accurately predict natural substances even when the noise percentage is high. Hence the signal-to-noise ratio was not an issue for the three tools, as proved by the lack of correlation between poor repeatability for white tiles and better results for apple TSS.
Considering the consistent results over tools and years, the scientists concluded that aging over three years had not impacted the predictive performance of any of the three tested SWNIR spectrophotometers. However, white tile was not a good reference material for apple TSS testing. The slight variation in performance lay in the visible spectrum range, which could and should also ideally be stabilized.
Testing the Predictors
The importance of studies like the one discussed cannot be over-emphasized, as entire supply chains of fruits and vegetables use these NIR instruments and their successors. While users should always be sure to maintain their devices according to manufacturer specifications, stakeholders can rest easy knowing that the measurements they make are still accurate, even over time.
- F-901 AccuRipe & AccuStore
- F-751 Melon Quality Meter
- F-751 Kiwifruit Quality Meter
- F-750 Produce Quality Meter
- F-751 Avocado Quality Meter
- F-751 Mango Quality Meter
- F-900 Portable Ethylene Analyzer
- F-950 Three Gas Analyzer
- F-920 Check It! Gas Analyzer
- F-960 Ripen It! Gas Analyzer
- F-940 Store It! Gas Analyzer
- Spectrophotometry in 2023
- The Importance of Food Quality Testing
- NIR Applications in Agriculture – Everything…
- The 5 Most Important Parameters in Produce Quality Control
- Liquid Spectrophotometry & Food Industry Applications
- Active Packaging: What it is and why it’s important
- Ethylene (C2H4) – Ripening, Crops & Agriculture
- Guide to Fresh Fruit Quality Control
- Melon Fruit: Quality, Production & Physiology
- Understanding Chemometrics for NIR Spectroscopy